Pastest's Tutor mode will happily hand you the answer. This guide is about not letting it — turning that AI layer into something that interrogates your reasoning for MRCP Part 1 instead of short-circuiting it. If you want the underlying reliability audit, read the grounding and hallucination audit; this is the implementation companion, built around twenty-five reusable prompts and one iron rule.
The rule that makes the rest work
Commit before you consult. Write your answer and a one-line rationale before Tutor mode opens, every single time. This is the anti-answer-leak rule, and it is the difference between practice that builds retrieval strength and practice that quietly replaces it with recognition. The learning science is settled: testing yourself and then getting feedback beats reading an explanation you were handed before you tried. An AI tutor available on tap makes breaking this rule effortless, which is exactly why the rule has to be explicit.
How to use the prompt library
The exam is two three-hour best-of-five papers weighted heavily towards clinical sciences and pharmacology; the distractors are engineered to be nearly right. So the prompts below are built to make the tutor expose the discriminating feature between options, verify the jurisdiction and date of its guidance, and generalise each miss into a transferable rule. Keep them generic — never paste proprietary question text into a prompt you reuse — and swap the specifics in as needed. They are grouped by the six error types from our workflow, because the right prompt depends on how you got it wrong.
Mechanism and knowledge-gap prompts
- "Without giving me the answer, what single mechanism distinguishes the correct option from the one I chose?"
- "Explain the underlying physiology in three sentences, then ask me a question that checks I understood it."
- "What is the highest-yield fact in this topic for MRCP Part 1 specifically, and why is it examinable?"
- "Give me the concept behind this item as a rule I could apply to a different presentation."
- "What would I need to know to answer any question on this topic, not just this one?"
Discriminating-feature and option-elimination prompts
- "For each of the five options, state the one feature that would make it correct — do not tell me which is right."
- "Why is my chosen distractor wrong? Name the specific discriminator, not the general topic."
- "Which two options are most easily confused here, and what separates them?"
- "If you changed one detail in the stem, which option would become correct?"
- "Rank the options from most to least defensible and justify the ordering before revealing the key."
Guideline-verification and jurisdiction prompts
- "What guideline or source supports this answer, and what is its date?"
- "Does this answer reflect current UK practice specifically? Flag any US/UK divergence."
- "Has the relevant guidance changed in the last two years? If you are unsure, say so."
- "Is this a settled recommendation or a contested one? Explain the disagreement if any."
- "Give me the source I should check to verify this independently."
Counterfactual and depth prompts
- "What is the most examinable exception to the rule this question tests?"
- "Construct a harder version of this concept and quiz me on it."
- "What would make this a two-step rather than one-step question?"
- "Which comorbidity would change the correct answer, and how?"
- "What is the classic trap examiners set on this topic?"
Retrieval-testing and consolidation prompts
- "Ask me three spaced questions on this topic over this session without repeating the original."
- "Turn my error into a single flashcard: cue on one side, rule on the other."
- "Summarise what I got wrong as one incorrect rule and one corrected rule."
- "Give me one transfer question on this concept in a different clinical context."
- "At the end, list the concepts I still could not explain unprompted."
The misconception record
For every genuine miss, write four lines: the incorrect rule you were using, the corrected rule, one transfer question, and a review date. This takes ninety seconds and is worth more than the prompt exchange that preceded it, because it converts a fluent explanation into a testable, dated commitment. Prompts 22–23 are designed to help the tutor draft it; you keep the record.
The weekly verification sample
Once a week, take five tutor outputs at random and check them against official or primary sources — the MRCP blueprint, a named guideline, the SmPC for a drug claim. Log every discrepancy. This is the same discipline the audit article formalises, and it is what lets you keep using Pastest's AI with calibrated rather than blind trust. Where you want a fast, cited second opinion on a UK-guideline point, Ask iatroX is built to return one with the source attached.
A seven-day pattern for busy trainees
Monday and Tuesday: 40 Pastest questions each, commit-first, prompts 1–10 on misses. Wednesday: a timed, unseen 50-question block in iatroX's free MRCP Part 1 bank — no tutor, pure measurement. Thursday: misconception-record review, prompts 21–25 for retrieval. Friday: 40 questions, prompts 11–15 to pressure-test jurisdiction on management items. Saturday: verification sample plus a second timed block. Sunday: rest. The prompts make Pastest interrogate you; iatroX measures you on unseen material; the misconception records are the durable output.
A worked prompt exchange
To see the difference between requesting an answer and testing reasoning, compare two ways of handling the same miss — a question where you chose "sarcoidosis" and the key was "berylliosis".
The lazy exchange: "Why is the answer berylliosis?" The tutor obliges with a paragraph on berylliosis, you nod, and nothing transfers — you have read a fact you will not be tested on again in that form.
The reasoning exchange uses the prompts in sequence. First, prompt 6: "For each option, state the one feature that would make it correct — don't tell me which is right." Now you are forced to articulate that berylliosis and sarcoidosis are radiologically and histologically near-identical, and that the discriminator is occupational exposure. Then prompt 11: "What source supports this, and what is its date?" — because if the tutor cannot anchor the exposure history's diagnostic weight, you have found a grounding gap. Then prompt 16: "What is the most examinable exception to the rule this tests?" — surfacing that a negative exposure history does not exclude the diagnosis. Finally prompt 22: "Turn my error into a single flashcard." You leave with a rule ("non-caseating granulomas plus occupational metal exposure → think berylliosis before sarcoidosis"), a dated source, an exception, and a spaced-review card — from the same miss the lazy exchange wasted.
The whole exchange takes about four minutes and produces something the exam can actually reward: a transferable discriminator, not a memorised answer. Multiply that across a revision period and the compounding is the entire point of the method.
Continue, supplement, switch or stop
Continue while commit-first holds and your verification log stays clean. Supplement with unseen timed blocks whenever the tutor feels productive but your measured performance is flat. Switch only on logged failures. Stop opening the tutor during blocks in the final fortnight — rehearse the assistant-free exam.
Frequently asked questions
Is Pastest enough for MRCP Part 1 on its own? Its volume and AI explanations can anchor preparation, but only if the anti-answer-leak rule is enforced and readiness is measured on unseen timed material outside the bank.
Which MRCP Part 1 component does Pastest not reproduce well? The unseen, blueprint-weighted, timed conditions of the real papers — and an always-available tutor actively threatens the retrieval practice those conditions demand unless you commit first.
How should I verify Pastest AI answers for MRCP Part 1? Use prompts 11–15 to force a named, dated source, check it directly or via Ask iatroX, and run a weekly five-output audit against primary sources with a written log.
When should I stop using Pastest and move to mixed mocks? Once coverage is complete, first-attempt accuracy is stable and pacing is on target — give the final two to three weeks to timed blocks with the tutor closed.
How should I combine Pastest with iatroX without duplicating practice? Pastest for commit-first drilling and reasoning prompts; iatroX for unseen measurement, Socratic repair and cited verification — the same question never appears twice.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 19 July 2026; Pastest features are vendor-published and evolving — verify on the product page. Disclosure: iatroX operates a competing MRCP Part 1 bank and Socratic Tutor. Corrections via the feedback route on iatrox.com. References: MRCP(UK) Part 1 format (thefederation.uk); Pastest MRCP pages (pastest.com); related reading: the Pastest MRCP AI audit and why your Q-bank percentage is not your exam score.